Chemical Engineering and Processing, 31 (1992) 91- 1 I7 91

Modeling and simulation of bubble column reactors

S. Schliiter, A. Steiff and P.-M. Weinspach

Lehrstuhl fir Therm&he Vecfahrenstechnik und W&me- und Stoffaustausch, Abteilung Chemietechnik, Unitiersitiit Dortmund,
Postfach 500 500, D -4600 Dortmund 50 (Germany)
Dedicated to Prof. Dr.-Zng. A. Vogelpohl on the occasion of his 60th birthday
(Received October 21, 1991; in final form December 9, 1991)

Abstract

The mathematical equations and a comprehensive computer program for modeling and simulating bubble column
reactors with internal heat exchanger installations are presented. Detailed reactor model equation systems on the
basis of the axial dispersion model and the cell mode1 with backflow are developed by linking together the
mathematical description of the multiphase flow properties with the momentum, mass and energy balances of the
reactor. A highly modular constructed computer program is presented summarizing a lot of theoretical and
empirical knowledge about multiphase flow and heat and mass transfer processes in bubble columns.
The enhanced possibilities of this new software system are demonstrated by simulating two industrial processes.
The methanol synthesis in the slurry phase is chosen in order to test the model accuracy for a multicomponent
reaction system including a thermodynamically balanced chemical reaction. Secondly, the wet air oxidation of
municipal sewage sludge in a large-scale slurry bubble column reactor is simulated as an example of a highly
exothermic oxidation carried out at extreme pressure and temperature conditions. This chemical process requires
special attention for reaction heat removal.
The numerical treatment of the algebraic equation system based on the cell model with backflow is easier and lcads
to average computing times up to 100 times lower than for the reactor mode1 based on the axial dispersion model.
The iteration process for solving the boundary value problem runs into numerical convergence difficulties for the case
of highly exothermic process conditions with considerable temperature gradients in the reactor. So the finite-difference
technique cannot be recommended from a numerical point of view, whereas the cell model formulation allows
computations on an IBM-compatible PC (386, 25 MHz) with tolerable computing times of the order of minutes.

1. Introduction-reactor types and industrial use fluid dynamics and special internal installations. Some
basic types are shown in Fig. 1.
In the chemical industry, bubble column reactors are A first classification can be made considering the
often used to carry out gas/liquid or gas/liquid/solid main liquid flow in the apparatus:
reactions [l, 21. In its simplest design the bubble - basic bubble column type without internal installa-
column is a vertical tube without any internal installa- tions or forced liquid circulation;
tions. In order to support gas/liquid phase separation - cascaded bubble column with limited gas and liquid
by decreasing superficial velocities, the reactor diameter backmixing;
is often increased in the upper part of the column. As a _ bubble column with internal loop (loop reactor,
gas sparging system, annular showers or simple sieve forced liquid circulation flow by hydrostatic pressure
trays can be used, for example. The dispersing energy is differences);
fed only by the gas phase, so further mixing and _ bubble column with external loop (forced liquid
moving of the fluids in the column occur only by their circulation flow by pump energy).
natural dynamics. Many gas/liquid reactions are connected with heat
Because operation of a bubble column does not production and therefore the removal of reaction heat
require any additional mechanically moving mixing is an important aspect in the design of bubble column
devices or similar internals, the bubble column reactor reactors in order to ensure safe process operations. In
can be considered as standard industrial equipment for Fig. 2 some examples of indirect heat transfer carried
multiphase reactions, with a high service life and oper- out in two- and three-phase bubble columns are demon-
ating safety. In industrial practice, because of special strated. Gerstenberg [3] and Steiner [4] report in detail
process engineering problems, a large number of bubble on the industrial use of bubble column reactors. Some
column types are in use, differing in the predominating examples are shown in Table 1. Organic syntheses with

0255-2701/92/$5.00 0 1992 - Elsevier Sequoia. All rights reserved

98

Z condenser

cooler

conventional cascade bubble column bubble column

bubble bubble with internal with external
column column loop loop

Fig. 1. Basic bubble column types used in industrial practice.

hydrogen, oxygen, sulfur dioxide and various olefines (1) description of the fluid-dynamic flow conditions
are the main applications. At present, the size of in- by consideration of liquid circulation and backmixing
stalled equipment ranges from 0.1 m3 up to 8300 m3 for properties with a fluid-dynamic model,
the BIOHOCH reactors for wastewater treatment. (2) development of a reactor model by linking to-
As can be seen from Table 1, the conventional bubble gether the fluid-dynamic description and the mass and
column without any additional installations is not used heat balances of the bubble column reactor under spe-
very often as industrial equipment. Cascading and cific boundary conditions.
forced liquid circulation flow are the most commonly The primary target quantities of a reactor model are
applied methods to solve special process operating the technical conversion data of the reacting compo-
problems. Heat removal is realized in most cases by nents. In some cases the thermal balance or the temper-
external liquid circulation. Owing to comparatively in- ature field in the reactor can be points of interest, too.
complete design data at present, internal longitudinal- In addition, a fully developed reactor model allows a
or cross-flow tube bundle heat exchangers are used only process to be optimized by analyzing the influence of
in exceptional cases. special adjustable parameters.

2.1. Fluid-dynamic models for bubble column reactors

2. Fluid dynamics and reactor models for bubble columns In Table 2, the most important fluid-dynamic models
for describing gas and liquid flow patterns in bubble
The mathematical description of the bubble column columns are briefly summarized. The basic physical
reactor can be interpreted as a two-step process: approach and the target quantities are shown for
 e
heat
corriel

heat
carrier

---- liquid
-----
liquid

external jacket cross longitudinal

heat heat coil flow flow
exchanger exchanger tube bundle tube bundle
Fig. 2. Heat removal methods carried out in two- and three-phase bubble column reactors.

comparison. For detailed information about the meth- cause problems if the strong nonlinear temperature
ods see the references or the comprehensive review in dependences of some model parameters are to be con-
ref. 38. The most important fluid-dynamic quantity is sidered in the general nonisothermal case. In addition,
the liquid circulation velocity field in the bubble some of the basic model assumptions or model parame-
column. Several other process engineering variables like ters have not yet been investigated experimentally.
the axial backmixing properties or the heat and mass
transfer coefficients can be correlated in relation to this 2.2. Reactor models for bubble columns
value. So the circulation flow models of Ueyama and As mentioned above, a reactor model is developed by
Miyauchi [ 18,391, Joshi and Sharma [23,24] and Zeh- linking the fluid-dynamic assumptions with the mass
ner [25,26] are widely used to estimate this velocity. and energy balances of the multiphase system. The
The axial dispersion model (ADM) given at the head theoretical concept and the physical and chemical quan-
of Table 2 cannot be regarded as a model describing the tities involved in this process are shown schematically
special bubble column fluid dynamics on a physical in Fig. 3. Selected works on modeling bubble column
basis. From a physical point of view the ADM is a fully reactors for industrial processes are summarized in
empirical model giving the unknown fluid-dynamic Table 3. The fluid-dynamic models mentioned in $2.1
properties of the multiphase system in the value of an have been used for describing the chemical reaction
‘axial dispersion coefficient’ as a lumped parameter. engineering data of bubble column reactors only in
Nevertheless, the dispersion model can predict gas- and exceptional cases. In particular, the very comprehensive
liquid-phase residence time distributions with an accu- multicomponent models proposed for describing the
racy sufficient for most technical cases (for example, see Fischer-Tropsch and methanol synthesis in the slurry
Table 3). phase allows us to conclude that the axial dispersion
Recently, some more physically based models have model is a good pragmatical basis for the development
been published by Gasche [20] and Torvik and Svend- of reactor models for industrial processes carried out in
sen [37] which can predict the gas/liquid-phase holdup bubble column reactors.
and the circulation velocity profile of both phases with It should be noted here that the cell model with
good accuracy. However, the mathematical equation backflow, which is widely applied in industrial prac-
systems arising from these approaches are of partial tice too (see Table 3), is a finite-difference equivalent
differential type and their numerical treatment will of the axial dispersion model. Considering specific
100

TABLE 1. Examples of the use of bubble column reactors in industrial practice

Product/process Material sources Basic type of reactor Heat removal method Refs.

Acetaldehyde Ethylene, oxygen BCR” without internal Direct vaporization of t3>51

installations product/water
Acetic acid Acetaldehyde, oxygen BCR cascaded External HEb [41
Acetic acid, methyl. n-Butane, air BCRs connected in cascade Longitudinal flow tube bundle [X 61
ethyl, ketone
Chlorinated paraffin n-Paraffin, chlorine BCRs (glass) connected in Internal cooling [31
cascade
Coal liquefaction Hard coal, hydrogen High pressure BCR (300 bar) Feed preheating [7,81
without internals or cascade
Cyclohexane Benzene, hydrogen BCR with external loop HE in external loop [31
Dimethyl terephthalate p-Xylene, oxygen BCR with internal loop 141
Ethylbenzene Benzene, ethylene BCR with external loop HE in external loop [31
Fischer-Tropsch Hydrogen, carbon monoxide BCR without internals External HE, longitudinal-flow [9, 101
synthesis in slurry phase tube bundle
Hydrogen peroxide 2-Ethyl anthraquinone, air BCR with internals [31
Hydroxylamine Nitrate, phosphoric acid BCR without internals ill1
Methanol synthesis in Hydrogen, carbon BCR without internals External HE, longitudinal-flow [12, 131
slurry phase monoxide tube bundle
Oxidation of montan wax Montan wax, chromic acid, air BCR cascade, internal cascade Direct vaporization of water [41
(sieve trays)

Paraffin sulfonate n-Paraffin, water, oxygen, BCR with UV’ lamp and HE in external loop [31
sulfur dioxide external loop
Propylene oxide Peracetic acid, propylene BCR cascade, internal cascade Internal HE [31
(sieve trays)
Side-chain chlorination Alkyl aromatics, chlorine BCRs with external loop and HE in external loop [41
of alkyl aromatics UV lamp, connected in cascade
Single cell protein Methanol, ammonia, air BCR with internal loop Internal cooling [31
Steam cracking of Vacuum residue, high High pressure BCR without External HE [I41
vacuum residue pressure steam internals

Wastewater treatment Wastewater, air BIOHOCHd reactor [41

Wet air oxidation of Sewage sludge, air BCR without internals External HE [15 161
sewage sludge

“BCR = bubble column reactor; bHE = heat exchanger; “UV = ultraviolet; %egistered trademark of Hoechst AG.

mathematical relations between the basic parameters of bubble column reactors has been developed. In this
both models (details are given in $3), the cell model section the basic model considerations and the resulting
with backflow leads to nearly the same numerical re- model equation systems are presented. An insight into the
sults if the number of mixing cells is assigned high developed computer program and the possible options
enough. A main advantage of the cell model is its great for performing calculations is given afterwards to point
flexibility in modeling processes with concentrated and out the highly pragmatic sense of this software project.
unequally distributed mass and heat sources.
3.1. Basic model considerations
The mathematical description of the bubble column
3. Development of a comprehensive computer program reactor is based on two different fluid-dynamic ap-
proaches:
A comprehensive computer program for simulation of (1) the one-dimensional axial dispersion model
chemical processes carried out in two- and three-phase (ADW;
 101

TABLE 2. Fluid-dynamic models for bubble columns

Model Authors [refs] Main target quantity Brief description and remarks

Dispersion model Levenspiel [ 171 Axial (and radial) Approach analogous to Fick’s first law of molecular
concentration profile of diffusion with ‘dispersion coefficient’ D,, as global
component in multiphase (lumped) model parameter:
system

Equation holds for both gas and liquid phase by

introducing respective phase holdup E

Recirculation flow models Ueyama and Miyauchi Liquid velocity profile, axial Starting point is force balance for differential
11% 191 dispersion coefficient cylindrical volume element:
a11
with T = -v,pL-
ar
Solution is given by introducing a special radial gas
holdup distribution, e.g.

g+, -$p) with +=g

and integrating over column radius with boundary

conditions given in ref. 18

Gasche [20,21] Liquid velocity profile, gas Approach similar to ref. 18 with force balance for
velocity profile both gas and liquid phase; the two balances are
connected by an additional force balance for a rising
bubble considering interaction of gas and liquid

Energy balance method Whalley and Davidson Liquid velocity profile Overall balance considering ‘hydraulic jump’ at free
WI liquid surface (for details see ref. 22); model equations
are limited to columns with HR/DR z 1

Circulation cell model Joshi and Sharma Mean circulation velocity of Enhanced energy balance method with multiple
[23,241 liquid phase circulation cells arranged in axial direction; circulation
+ axial dispersion coefficient cell height determined empirically to be Hc x 0.8 D,
4 heat transfer coefficient

Cylindrical vortex Zehner [25,26] Mean circulation velocity of Liquid circulation is given by gas holdup difference
model liquid phase between two circulation cells arranged in parallel:
-+ axial dispersion coefficient
+ heat transfer coefficient (p,_ - pG)gHv AB = [ 2 v~! with BE z$ , Hv x D,
+ mean gas holdup c
Quantity AE‘is given as function of bubble rise
velocity and circulation velocity; model parameter i is
estimated empirically as c = 2.5

Bubble class models Joseph and Shah Gas-phase residence time Two bubble classes are assumed: large bubble class
[27,281 distribution in plug flow and small bubble class perfectly
mixed; no interaction between small and large
bubble strand; separated mass balances set up for
both classes

Grund [29] Gas-phase residence time Similar approach to ref. 27 but considering
distribution coalescence interaction between small and large
bubble strand; theoretical model, at present no
experimental background to determine local
dependent coalescence functions

Continued overleaf
102

Table 2 continued

Model Authors [refs] Main target quantity Brief description and remarks

Tank-in-series model Levenspiel [ 171 Residence time distribution of Perfectly mixed reactor cascade without backflow
(cascade model) gas and liquid phase between neighboring cells; model assumes equal
cell numbers (and therefore equal backmixing
properties) for gas and liquid phases

Cell model with Popovif [30, 311 Residence time distribution Perfectly mixed reactor cascade with backflow between
backflow Mecklenburg and of gas and liquid phase neighboring cells; the two model parameters, r or Y
Hartland [32], and NC, can be linked theoretically with the axial
Roemer and Durbin dispersion model:
[331
NC 1 UC&
r=--- with Boo=-
Boo 2 gGDsff, o
NC 1 ULLR
Y =~o,-$ with Bo,=
G_eef. I_
In this formulation, the cell model with backflow is a
finite-difference equivalent of the axial dispersion
model outlined above
Slug cell model Myers et al. [34,35] Various properties of gas See refs. 34 and 35
and liquid phases
Enhanced cell models Wasowski and Blat3 Various properties of gas Several theoretical models for description of large
[361 and liquid phases and small bubble behavior considering coalescence
and dispersion mteraction between phases; at
present, experimental verification is not possible
Fluid-mechanical model Torvik and Svendsen Distribution of phase Comprehensive fluid-mechanical description of bubble
[371 velocities and mass column reactor; for details see ref. 37, the partial
concentrations of all differential equation systems are solved numerically by
components several complicated algorithms

(2) the cell model with backflow between neighbor- cocurrent gas/liquid JEow operation
ing cells (CWB).
~Ci,=!&= EGDeR,G/(d o ug’) - 6(J)/2
These classical one-parameter models are chosen
VG (1 + 6”‘)/2
from the approaches given in Table 2 because their
simple and clear construction lead to less numerical and
effort in solving the resulting balance equation systems.
(i,_ i/,,,._~~De~&o~ij)) -S”‘l2
Some of the more physically based fluid-dynamic mod- Y
JL (1 + 6”‘)/2
els can easily be involved by using the resulting equa-
tions for the mean gas holdup ao, gas- and liquid-phase countercurrent gas/liquid flow operation
dispersion coefficients D,, G and DeK,L, and the heat
transfer coefficient c+. given by the authors in their
relevant papers (e.g. Ueyama and Miyauchi [ 181, Joshi
and Sharma [23, 24,481, Wachi et al. [ 19, 391, Zehner and
[25,26]). For the development of an easy-to-use soft-
(j,=&=GD~Ir,.l(d@UL (j)) - l/2
ware tool this method seems to be a good compromise Y
PI. ( 1 + P)/2
between a realistic view of the fluid dynamics and a
possibly low computing effort. semi-batch gas/liquid JIow operation
The model parameters of the CWB can be connected
&Den L 1 + 6”) (jj _ CDen- L
with the axial dispersion coefficient of the ADM in order d,=2-=-L LY
to transform the CWB into a finite-difference equivalent A,

of the ADM. Under consideration of the axial depen- with

dent cell heights the following equivalence equations can A, _+&i)+@ifl)
2 c c 1
be derived [38] from a Taylor expansion for the molar
concentration c$j) of a component i in the cell (j): and
Fig. 3. Schematic reactor model concept.

High pressure vapor, water or various heat carrier

fluids given in the VDZ- Wtirmeatlas [ 501 can be used for
heat removal. Material data and heat transfer co-
On the basis of these equivalences the following
efficients are computed locally as functions of pressure,
model considerations hold for both the ADM and the
bulk fluid and wall temperatures. The gas and liquid
CWB formulation.
phase are assumed to have the same temperature.
(1) The basic mathematical decription follows the
(6) Thermal backmixing is completely given by
axial dispersion model (ADM) and the cell model with
backmixing of mass:
backjlow (CWB). The equation systems are designed as
a multicomponent model considering parallel and con- 5, = EG’&, G + ( 1 - ~Gkt-, 1
secutive reactions in the liquid (slurry) phase.
= EGPGC~, GDem G + ( 1 - &Ghcp. L&T, I.
(2) The bubble column reactor can be operated in
cocurrent, countercurrent or semi-batch flow operation (7) Axial dependent reactor pressure is considered;
mode. there is assumed to be no pressure difference between
(3) The multiphase system is assumed to be in a the gas and the liquid phase.
stationary state. (8) Axial dependent superJcia1 gas velocity is consid-
(4) The axial dependent solid concentration can be ered, taking into account gas/liquid mass transfer and
estimated by the sedimentation-dispersion model in the expansion effects.
formulation given by Smith and Ruether [49] assuming (9) The overall mass transfer resistance can be as-
a monodispersed solid phase. sumed to be completely on the liquid side (kL = &J; the
(5) A comprehensive heat transfer model is integrated interphase concentration of a component can be ap-
considering cross-flow longitudinal-flow tube bundle proximated by Henry’s law.
heat exchangers and heat transfer to the reactor wall. ( 10) Gas/liquid mass transfer with chemical reaction
104

TABLE 3. Selected papers on modeling bubble column reactors

Authors [refs.] Year Backmixing model Chemical reaction Remarks

gas/liquid/solid

PopoviE [ 30, 3 1] 1973175 CWB”ICWBI- Absorption of isobutene in Axial dependent superficial gas
sulfuric acid velocity
Chianese rl u/. [40] 1981 PFMb/CWB/- Alkylation of benzene NC = 10 cells, multifeed
with ethylene
Deckwer et al. [41] 1982 ADM’/ADM/ADM FTd synthesis in liquid Balanced component: H,
phase
Brown ef al. [42] 1983 TIS=/CWB/CWB Methanol synthesis in Experimental data: D, = 60 cm,
liquid phase La = 5.0 m
Kuo et al. [lo] 1983 PFM/ADM/- FT synthesis in liquid Balanced components: H,
PFM/ADM/- phase Balanced components:
H,, CO, CO,. H,O
Stem et al. [43] 198.5 ADM/ADM/ADM FT synthesis in liquid Balanced components: Hz, CO,
phase CO,, H,O, HC?
Joseph er al [28] 1985 Two bubble class model Absorption of CO2 in Da = 10.2 cm, La = 2.36 m
NaOH solution
Schumpe et al. [44]
Myers et al [3S] 1987 Slug and cell model Model reactions of first D, = 0.19/O. 14 m, L, = 2.40 m
and second order,
absorption of Oz in water
Wachi and Morikawa 1987 PFM/CWB/- Chlorination of ethylene Reaction heat removal by direct
1391 vaporization of liquid product;
model with integrated energy
balance
&tiirk ei al. [45] 1988 ADM/ADM/ADM Methanol synthesis in Heat balance integrated
liquid phase
Gasche [20] 1989 Two-dimensional dispersion Chlorination of toluene Constant reaction temperature
model with radial (slow reaction),
dependent circulation flow Absorption of CO, in
NaOH solution (fast
reaction)
Turner and Mills [46] 1990 ADM/ADM/ADM FT synthesis in liquid Balanced component: H,
phase
PFM/TIS/TIS, TIS/TIS/TIS Empirical: NC = OSO(L,/D,)‘-’
Krilik et al. [47] 1990 -/ADM/- Reaction of hydrogen, Dynamic model
acetone and
4-aminodiphenylamine
Torvik and Svendsen [37] 1990 Comprehensive Methanol synthesis in
fluid-mechanical model liquid phase

“CWB, cell model with backflow; bPFM, ideal plug flow model; “ADM, axial dispersion model; dFT. Fischer Tropsch process converting
hard coal to hydrocarbons; ‘TIS, tank-in-series model; ‘HC, representative hydrocarbon component.

is considered, assuming the validity of the film theory. model are summarized in Table 4. The equation system
( 11) All fluid-dynamic parameters (DeE, =, D,, G, results from a balance for a differential volume element
k,a, c+, w,., . . . ) and the material data
2eff,GT /2,, L, EGG, of the bubble column reactor. Following the consider-
for the gas and liquid phases are correlated locally for ations described above, an ordinary boundary value
each cell and net point as functions of local pressure, problem is stated consisting of two first-order nonlinear
temperature, internal installations and superficial phase differential equations
velocities. - momentum balance, liquid phase
- overall mass balance, gas phase.
3.2. Model equations for the axial dispersion model two second-order nonlinear differential equations for
The basic model equations of the axial dispersion each component balanced in the gas and the liquid phase
 105

TABLE 4. Differential equation system for the axial dispersion TABLE 5. Boundary conditions for the axial dispersion model
model

PO-) = P(l+). %(I-) = 0, %(L_) = 0. S(l_) = Jc*(o-). %(I-) = cl

conntc-“l gauuqldd oorr mniilionc

maw balance, component i, liquid phm:

&.p-(,,,,_ caDa.a &-)@I+) = 0

(c&o+) = 0

.
+~~w,,.w,w(Tn-Tw,.,)
I+=1 I
(z)(o+, = 0
A. sLm,~dT”
(ifrs.pTo)(0-)- (&cp,~l;) CO+)+ (O+)= 0( d=)
$(*-) = 0,

_ mass balance component i, gas phase

_ mass balance component i, liquid phase
the same as described above for the axial dispersion
and one second-order nonlinear ordinary differential
model in $3.2. Cell arrangements and flow directions
equation describing the differential energy balance of
for cocurrent, countercurrent and semi-batch flow con-
the multiphase system. Upstream and downstream
ditions are given in the Appendix. The (Nc x N,,)
boundary conditions for continuous flow systems are
equation system, resulting from these arrangements and
discussed in detail by Novy et al. [51]. Here, only the
the considerations of @3.1, is of nonlinear algebraic
classical ‘Danckwerts boundary conditions’ first given
type. A detailed specification is given in Table 6. The
by Danckwerts [52] are used, which state a jump in
mass and heat sources and transfer terms of a cell (j)
axial backmixing at the reactor boundaries from zero to
can be written as follows:
its uniform nonzero magnitude within the reactor. In
Table 5 the boundary conditions are summarized for
cocurrent, countercurrent and semi-batch flow opera-
tion. The symbols (-) and (+) signify locations imme-
diately upstream and immediately downstream of a
N
plane perpendicular to the mean flow at the reactor
bottom (0) or the reactor head (1).

3.3. Model equations for the cell model with backflow &“=H~‘A,(l -e&i) f Rk(-AHrqk)
k=l
The model equations for the cell model with
backflow result from N, overall momentum, mass and The reactor boundary conditions are included by bal-
energy balances for each cell (j) inside the bubble ancing cell (1) and cell (Nc), taking into account the
column reactor. The Neq quantities to be balanced are specific inlet and outlet conditions of the column.
106

TABLE 6(a). Algebraic equation system for the cell model with TABLE 6(b). Algebraic equation system for the cell model with
backflow (part I) backflow (part 2)

~ss~~-mmponcn'i.~uidph.rc,mrmhlcum ntgasmquid
Bow
j=*:
f&w _ $1'
(1 + 7"))("l,c,,L)'2)-
(1+7'~))(",Q.)"~+
-%+=O
l<j<Nc:

(l+,fJ))(",c,,)"+"- (1+."-1)+70))(ULCi,L)G'f

‘+“) _ $‘I
(“L,, c*J2J- (UL,,Q.)“) + y =0

(ILL,, e.p - 2 (“L,,c,,L)“’+ (UL,, .JJ+‘) + %y$ = 0

j=N,:
(%.1cd

j= 1:

+c’~‘(upcplp)~’
- (1 +F'q(upc,o)~'
t
1 < , < NC :

!$? (,&‘) _ ,+‘) + (l,,LpL)‘J’ (,+I) - ht") + (7uLpL)“-‘) (,-‘I - hl") +

+ E”)(w,b), 6+‘) _ (1 ++I ++-‘))(upc,d):)+ (I + f”-‘))(upe&‘)+
‘I’
J = NC :

3.4. Survey of the developed computer program

To consider the complicated nonlinear interrelations various fluid-dynamic quantities in different flow opera-
and dependences of the various model quantities out- tion ranges and fluid types, are implemented using the
lined in Fig. 3, the developed program constructed is information given in the literature. To simulate a spe-
highly modular. The special software modules are cific problem the user has to supply an input data set
designed to summarize the present scientific knowledge containing information about geometrical reactor
about scales, material specifications, operation conditions, re-
_ the fluid-dynamic properties such as gas holdup, fluid action kinetic specifications, heat transfer methods and
circulation velocities, mass and heat transfer coefficients specifications about the methods to be used by the
as functions of local temperature, pressure, material program for calculating the necessary parameters de-
properties and operating conditions specified by the scribing phase holdups, circulation velocities, gas- and
user; liquid-phase backmixing, heat and mass transfer pro-
_ the heat transfer operations including jacket cooling cesses, and several other quantities less important here.
and cross-flow or longitudinal-flow tube bundle heat The nonlinear differential or algebraic equation sys-
exchangers with various heat carrier fluids; tems given in $8 3.2 and 3.3 cannot be solved analyti-
- the temperature and pressure dependences of the cally and are therefore to be treated by numerical
various model quantities; and methods. The module D02RAF from the NAG Fortran
- the influence of reaction kinetics on gas/liquid mass Library is used for solving the nonlinear boundary
transfer enhancement effects. value problem of the reactor model based on the axial
For this purpose a large number of semitheoretical dispersion model. The numerical background of this
and empirical parameter correlations, developed for routine is a finite-difference method, using a deferred
 107

TABLE 6(c). Algebraic equation system for the cell model with program version running on a widely used low cost
backflow (part 3) computer system. The module HYBRD from the MIN-
PACK Library, which is fully equivalent to the routine
COSNCF from the NAG Library, is used as the equa-
tion solver module in this case.
To give an insight into program structure, an extract
from the program flow scheme is shown in Fig. 4. The
module FCN is designed to determine the balance
I<j<Nc: residuals for the equation system based on the cell
model with backflow and assign it to a residual vector,
which is analyzed by specific numerical software mod-
ules in order to calculate a new estimation for the Neq
unknown model quantities. For this purpose first the
complete fluid-dynamic and thermal description for
cells .j = 1,. . , N, has to be stated. This is done by
calculating the necessary material, fluid-dynamic, ther-
modynamic and kinetic data of each cell and combining
them with specific transfer terms. After this task is done
the residual vector can be determined simply by reading
the model equations given in Table 6 in terms of the
specific programming language.

l<j<N,:
4. Methanol synthesis in the slurry phase
(q,,p#’ (hk+‘l hk’) + (uL,,p#-” (hl'-"-
hk') t
t ~“‘(Wc,qO
6+I)- (1 + [(‘I + F~J-~))(upe,ff)l;‘)
+ (1 + F(‘-~))(“pe,ff)“-‘)
G
+ Methanol from synthesis gas is one of the most
important bulk chemicals with a world production of
about twenty million tonnes per year [56]. The main
chemical equilibrium reactions of the catalytic hydro-
genation of carbon monoxide,

CO + 2H, Z$ CH,OH

(AH’&, = - 90.84 kJ/mol)

correction technique and Newton iteration [53]. The
nonlinear algebraic equation system arising from the CO* + 3H, + CH, OH + H,O
cell model with backflow is treated by the routine
COSNCF from the NAG Fortran Library, using a (A@,, = - 49.57 kJ/mol)
scaled version of Powell’s hybrid method [53--551.
COz+H,+CO+H,O
As an alternative to the finite-difference technique,
the use of a collocation method is possible. At present (AH& = -41.27 kJ/mol)
a program version is developed using the collocation
routine D02TGF from the NAG Library [53] to solve are exothermic in the direction of alcohol formation. As
the system of nonlinear differential equations after a new development in the field of methanol technology,
quasilinearization by a Taylor expansion. A compari- Chem Systems Inc. has proposed to carry out the
son of numerical results, computing times and conver- methanol synthesis in a slurry bubble column reactor
gence behavior is to be published later. with a catalyst suspended in an inert organic phase [ 571.
To have access to the NAG Fortran Library the A detailed comparison of the advantages and disadvan-
program has to be implemented in the programming tages of the fixed bed and slurry-phase methanol pro-
language Fortran 77 at the central computer of the cess is given by Oztiirk et al. [45]. The flow scheme of
computing center of the University of Dortmund. In the methanol three-phase process developed by Chem
the simulation runs low computing times and good Systems [58] is shown in Fig. 5.
convergence behavior of the program version based on The removal of the reaction heat of about 91 kJ/mol
the cell model with backflow were obtained. So it carbon monoxide requires special attention. Chem Sys-
appeared useful to implement this method on an IBM- tems Inc. proposed external liquid circulation including
compatible PC in order to get a portable stand-alone a heat exchanger for producing high pressure steam.
108

Fig. 5. Basic flow scheme of the simulated methanol slurry

process.

The reactor temperature can be controlled by the satu-

ration vapor pressure in the tubes. An alternative way
of heat removal is an internal tube bundle heat ex-
changer of cross-flow or longitudinal-flow type. In this
case external fluid circulation is not necessary.

4.1. Basic considerations

The reactor model describing the methanol slurry-
phase process is developed as a multicomponent model
balancing CO, H2 and the main product CH,OH in the
gas and liquid phase. According to measurements of
Kotowski et al. [59], Sherwin and Blum [ 131 and von
Wedel et al. [60], the CO,-conversion reaction and the
water gas shift reaction given above can be neglected
under the operation conditions proposed by Chem Sys-
tems ( < 10% CO, feed gas, suspended CuO/Zn catalyst).
The kinetics of the slurry-phase methanol formation
on industrial Cu/ZnO catalysts suspended in an inert
hydrocarbon have been investigated by Graaf et al.
[61,62], Ledakowicz [ 631 and von Wedel et al. [60]
using autoclave experiments. For the overall reaction
rate von Wedel et al. propose the following power law
expression which is used in our reactor model:

Rco = 8.79 x lo5 exp

~
CCZQ

The computer simulations are carried out on the

basis of the axial dispersion model (ADM) and the cell
model with backflow (CWB) given in @ 3.1 and 3.2.
The basic model equations are given in Tables 4, 5 and
6 with CO, H2 and CH,OH as balanced components
(i). The axial catalyst distribution can be derived explic-
itly for both models from the mass balance for the
monodispersed catalyst phase given in Table 4.

4.2. Selected computer simulation results

Fig. 4. Extract from the program flow scheme: setting up the
To give some examples of the computer simulation
balance residual vector (cell model with backflow). results, Figs. 6 and 7 illustrate the temperature and
 109

530 reactor temperature [Kj s8 CO conversion (gas phase)

528
:;I
!m
5.24 :,5 !

522 34

520 33

518
516
I
3.0 4.0 5.0 6.0 7.0 8.0 6.0 so
10’ IO’ 1 ’ ’ IC’ lo3
reactor height [m]
liquipphase’fspers’i~n coefficient[nk/s]
Fig. 6. Axial temperature plot for the methanol slurry process. Fig. 8. Influence of liquid-phase backmixing on the overall CO
conversion in the methanol slurry bubble column reactor.

4. CO - conversion (gas phase) [ss]

35 6. CO - conversion (gas phase) [%]

30

25

20

15

10

0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0
reactor height [m]
0 2 4 6 8 10 12 14 16 1s

Fig. 7. Axial CO-conversion plot fdr the methanol slurry process. mean superficial gas velocity [m/s]

Fig. 9. Comparison of reported and computed overall CO-con-

version data for the methanol slurry process.
TABLE 7. Basic data set for the methanol slurry process simula-
tions
of backmixing is not especially high due to a largely
Model parameter Quantity kinetically and thermodynamically controlled reaction
regime. The highest conversion is reached not for the
Column diameter l.Om
Column height 9.0 m plug flow model but in the middle range of the disper-
Reactor operating pressure 80.0 bar sion coefficient. In the case of plug flow, the liquid feed
Suspension feed temperature 187.0 “C is not heated up by thermal backmixing. This causes a
Mean superficial gas velocity 7.0 cm/s lower reaction rate in the lower part of the bubble
Mean superficial liquid velocity 1.O cm/s
column and results in a lower overall carbon monoxide
H&O feed gas ratio 2.0
15.0 vol.%
conversion. Straight after reaching the maximum, the
Inert gas fraction
Catalyst mass fraction 30.0 mass o/u conversion decreases due to the increase of backmixing
Particle Sauter diameter 30.0 pm and a resulting higher integral methanol concentration
in the reactor. For this reason the possible thermody-
namic equilibrium conversion is lowered by the law of
conversion plot for the bubble column reactor operat- mass action.
ing under the conditions shown in Table 7, cocurrent The model accuracy can be discussed by comparing
gas/liquid flow and external liquid circulation flow for simulation results with conversion data reported for a
reaction heat removal as shown in Fig. 5. The tempera- pilot plant operated by Chem Systems Inc. [ 131. Figure
ture increase between the reactor bottom and the reac- 9 is a plot of reported and computed overall carbon
tor head is about 12 IL monoxide conversion against the superficial feed gas
The influence of liquid-phase backmixing on the velocity. This plot is selected because the superficial gas
overall carbon monoxide conversion in the bubble velocity is used in nearly all fluid-dynamic parameter
column reactor is demonstrated in Fig. 8. The influence correlations and is therefore one of the most important
110

(1) components easy to oxidize (component family

IA))>
(2) components difficult to oxidize (component fam-
ily {B}), and
(3) components very difficult to oxidize or not oxi-
dizable (component family {C}).
For the oxidizable organic categories {A) and (B}
the authors report the following second-order kinetics
(the temperature dependence is correlated from the
0 hornme‘ mull e bmt ercha”p
data given in ref. 68):
b slurry tonl. f reOC,w
c hl$l pressure pimp ‘lwl tower 12137
d otr compressor
9
h Ilk R-LoU,A = 8.889 x 10~ 6 exp 27.781 -~
T >
Fig. 10. Basic flow scheme of the simulated wet air oxidation
process. x ‘COD. A cOl, L

6
10144
R COD,R = 8.889 x lo- exp 21.556 - ~
model parameters. A comparison with the pilot plant T >
data shows that the computed results are about 4%
x CCOD, B cOz,L.
lower than the measured ones. In addition, Fig. 9
indicates that the influence of the superficial gas veloc- Assuming a mean carbon/hydrogen ratio of x/y in
ity on the fluid-dynamic properties in the bubble the organic sludge components, the following overall
column reactor is predicted correctly by the model. model reaction can bc written for wet air oxidation:
The difference between computed and reported con-
version data can be explained by the solubility data of
carbon monoxide and hydrogen for the various waxes.
The Henry coefficients reported by von Wedel et al. In our model computations, the values x = 3 and
[64], which are used in this paper too, are up to 40% y = 4 are used. The heat of reaction is of the order of
higher than those measured by other investigators for 13 MJ per kg oxygen consumption. Beyrich [69] gives a
various paraffinic waxes. This results in a possible CO range of 12.6-14.7 MJ/(kgCOD). In our computer
conversion decrease of up to 10% and the conclusion simulations, a mean value of AH;,, = 13.6 MJ per kg
can be drawn that the waxes used by the Chem Systems oxygen consumption is used. In the classical Zimmer-
Inc. [ 13, 581 have a higher solubility for synthesis gas mann process, the reactor temperature is controlled
components than those used by von Wedel et al. in only by the sludge feed temperature without any addi-
their investigations. tional heat removal installations. Heat recovery is
therefore limited to the mass flows leaving the bubble
installations. With regard to better process temperature
5. Wet air oxidation of municipal sewage sludge control and energy recovery, two other methods for
reaction heat removal are possible:
In 1911, a process to oxidize sulfite waste liquor from ~--jacket cooling+ heat transfer to the reactor wall, or
pulp production in autoclaves at a temperature of about _ installation of an internal cross-flow or longitudinal-
180 “C had already been patented [65]. In the 1960s a flow tube bundle heat exchanger.
number of wet air oxidation plants for the treatment of Jacket cooling, however, cannot be recommended for
municipal sewage sludge were installed in the USA by the process conditions of wet air oxidation. A reactor
Zimpro Corporation, including a large-scale wet air wall thickness of about 100 mm and more causes a
oxidation process in Chicago [65] which is well known relatively high heat transport resistqnce which predomi-
because of its size and operating conditions. Figure 10 nates over the overall heat transfer coefficient: Thus,
shows the basic flow scheme of the continuous Zimmer- either the amount of energy transferred to the heat
mann process as described by Angenend [ 151. carrier fluid is low or the outlet temperature of the
cooling medium is relatively low, which prevents effec-
5. I. Basic considerations tive heat recovery.
Ploos van Amstel and Rietema [67, 681 investigated The internai tube bundle heat exchanger provides
the kinetics of the wet air oxidation of municipal large specific heat transfer surfaces and an extremely
sewage sludges. They suggest that the sludge compo- large overall heat transfer coefficient for reaction heat
nents should be divided into three main material cate- removal. Therefore it is possible to produce high
gories differing in overall oxidation reactivity: pressure steam at temperatures above 270 “C and simul-
 111

TABLE 8. Basic data set for the wet air oxidation process

Model parameter Quantity

Column diameter 2.0 m

Column height lX.Om
Column wall thickness 110.0 mm
Reactor operating pressure (head) 125.0 bar
Suspension feed temperature I 190.0 “C
Suspension feed temperature II 250.0 “C
Steam temperature (tube) 280.0 “C
Suspension mass flow rate
Overall COD value
72.0 tonne/h
60.0 kg/m3 33
0 2 L 6 8 10 12 1~ 16 m ‘s
Superficial gas velocity (bottom) 2.24 cm/s reactor axIs -
Superficial liquid velocity 0.8 I cm/s
Solid-phase mass fraction 3.0 mass % Fig. 11. Comparison of concentration profiles of component fam-
Mean Sauter diameter of solids 100.0 pm ily {A} obtained from the axial dispersion model (ADM) and the
cell model with backflow (CWB).

taneously control the process temperature safely at a

very high level. In addition the reactor can be preheated
by steam condensing in the tubes during the startup
phase.
Using the overall kinetic expression given above, the
reactor model is formulated as a four-component de-
scription summarizing the various organic sludge com-
ponents in two representative reactivity classes
indicated as {A) and {B). The basic model equations
are given in Tables 4, 5 and 6 with 0, and CO,
balanced in the gas and liquid phase, and sludge d 15nl
2647 e l.Om
families {A} and {B) which are, respectively, easy and
difficult to oxidize, balanced only in the liquid phase. 250
0 2 L 6 8 IO 12 IL 16 m 18
The axial dispersion model leads to a boundary value reactor height -
problem described by two first-order and seven second-
Fig. 12. Axial temperature plots for the wet air oxidation reactor
order nonlinear ordinary differential equations. For the with an external heat exchanger.
cell model with backflow, a system of nine nonlinear
algebraic equations per cell is to be solved numerically.
the order of NC = 10.
5.2. Selected computer simulation results To give an impression of the numerical computing
The basic process data summarized in Table 8 are required: the step curve for N, = 10 shown in Fig.
chosen as being similar to the process operation data 11 was computed on an IBM-compatible PC (386,
given by Angenend [ 151 and Randolf [66] for the wet 25 MHz, numerical coprocessor installed) with an over-
air oxidation plant installed in Chicago. In addition, the all computing time of about 100 s; in comparison, the
possible methods of reaction heat removal outlined computing time consumed by the ADM-based reactor
above are simulated assuming a longitudinal-flow tube model would be of the order of three hours, if the
bundle heat exchanger for producing high pressure simulations were carried out on the same computer
steam of 65 bar. system.
A comparison is given in Fig. 11 of simulation results The axial temperature plots for bubble column reac-
obtained from reactor models based on the axial disper- tor with external (suspension feed temperature 190 “C)
sion model and the cell model with backflow (overall or internal (suspension feed temperature 250 “C, steam
cell numbers: N, = 10 cells and 60 cells, respectively). It saturation temperature 280 “C) heat exchangers are
can be seen that for the operating conditions of the shown in Figs. 12 and 13. The curve parameters are the
highly exothermic wet air oxidation process the pro- inner reactor diameter or the suspension feed tempera-
posed cell model gives a nearly perfect copy of concen- ture, respectively. The reactor without an internal tube
tration profiles obtained from the ADM. Due to the bundle heat exchanger shows a temperature increase of
high degree of liquid-phase backmixing, the overall about 50 K up to an outlet temperature maximum
number of mixing cells can be chosen relatively low, of above 300 “C, whereas the temperature increase in the
 (1 220°C d 250%

reactor height -

Fig. 13. Axial temperature plots for the wet air oxidation reactor
with an internal longitudinal-flow tube bundle heat exchanger.
suspension 1.5MW

outlet
6.7MW Fig. 15. Overall thermal balance for the wet air oxidation reactor
fi hiah cressure steam
with an external heat exchanger for a sludge suspension feed
temperature of 190 “C.

energy recovery process depends on technical expendi-

ture. In comparison, a considerable amount of energy
can be recovered as high pressure steam by installing an
internal tube bundle heat exchanger. In addition, the
amount of energy recovered by preheating the entering
feed suspension is considerably higher in the case of
internal heat removal. This feature is caused by the
higher feed suspension temperatures necessary for high
organic component conversion. Further numerical sim-
suspension ulations indicate that an increase in the feed suspension
1.5MW
17MW
temperature causes an increase in the energy recovery
Fig. 14. Overall thermal balance for wet air oxidation reactor potential of both internal steam generation and feed
with an internal longitudinal-flow tube bundle heat exchanger for preheating.
a high pressure steam temperature of 280 “C and a sludge suspen- However, a simultaneous decrease in the temperature
sion feed temperature of 250 “C.
difference between feed and outlet flow increases the
technical effort for feed preheating. Therefore, an opti-
reactor with an internal tube bundle heat exchanger is mal solution has to be found, considering the effective
limited to nearly 8 K in the temperature range of 280- costs for feed preheating, the profit of high pressure
288 “C. From an optimal process control, installation steam generation, and a better process temperature
of an internal tube bundle heat exchanger is superior to control.
the classical method of external heat exchange between
the fluids entering and leaving the reactor.
The overall thermal balance for the bubble column 6. Conclusions and recommendations
reactor with and without internal cooling is demon-
strated in Figs. 14 and 15. All enthalpies refer to a The results of multicomponent nonisothermal bubble
temperature level of 0 “C. The mechanical power for column reactor modeling given in #$4 and S demonstrate
feed gas compression is also given. The energy distribu- the high flexibility of the developed reactor model based
tion for a specific case depends on the process operating on the axial dispersion model (ADM) and the cell model
conditions. Nevertheless, the figures give a good quali- with backflow (CWB). In most cases the model accuracy
tative view of the heat balance of the wet air oxidation depends less on the degree of mathematical complexity
process carried out in the bubble column reactor using of the fundamental fluid-dynamic model equation sys-
two different heat removal methods. tem than on the accuracy of specific coefficients describ-
To use the reactor without internal heat exchange, ing phase holdups, reaction rates or heat and mass
the hot gas/liquid mixture leaving the reactor must be transfer data in the multiphase system. So a complete
prepared for optimal thermal use. The effectivity of the process engineering description of the bubble column
 113

reactor (including internal and external forced circula- =F reactor length, m

tion flow and an extended heat transfer model) based A4 mass flow rate, kg/s
on more pragmatic attempts like the ADM and CWB is N overall number of components
a good approximation to the physical reality. N.C overall number of mixing cells
The developed computer program summarizes a lot N molar flow rate, mol/s
of bubble column research and gives the process engi- n numerical parameter
neer a good interface for simulation of bubble column p. pressure, kg/( m s*)
reactor operation with various enhanced options. At Q heat flow rate, W
present, the numerical treatment of the algebraic equa- 52 molar gas constant, J/(mol K)
tion system based on the cell model with backflow is R reaction rate, mol/(m3 s)
much easier and leads to average computing times up to r radial coordinate, m
100 times lower than for the reactor model based on the T absolute temperature, K
axial dispersion model, if a finite-difference technique is t time, s
used for solving the numerical problem. u superficial phase velocity, m’/( m* s)
The iteration process for solving the boundary value ‘G. < superficial gas backflow velocity, m/s
problem arising from the axial dispersion model runs u.L, Y superficial liquid backflow velocity, m/s
into numerical convergence difficulties for the case of V volumetric flow rate, m3/s
highly exothermic process conditions with considerable VG, t gas backflow rate, m3/s
temperature gradients in the reactor. So the finite-differ- VL: Y liquid backflow rate, m3/s
ence technique cannot be recommended from a numeri- V velocity, m/s
cal point of view. A program version using an V, liquid circulation velocity, m/s
orthogonal collocation method for solving the quasilin- Vs hindered settling velocity, m/s
earized differential equation system is being developed Vscc terminal settling velocity, m/s
at present in order to overcome the numerical disadvan- x axial coordinate, m
tages of the continuous gradient model. Z* gas/liquid flow operation indicator
The reactor model based on the CWB has been
implemented on an IBM-compatible PC (386, 25 MHz) c7 heat transfer coefficient, W/(m2 K)
in order to get a portable stand-alone program version Y liquid-phase backflow ratio
running on a widely used low cost computer system. A difference
The computing times required by this program are of AH%,, reaction heat (standard conditions), J/mol
the order of minutes, which is more than tolerable for A, mean cell height in forward direction, m
process engineering work. 6 backward/forward cell height ratio
E volumetric phase holdup
E overall volumetric phase holdup
Nomenclature Celsius temperature, “C
; parameter in Zehner’s model
A area, m2 Lr effective thermal conductivity, W/( m K)
A, B sludge reaction rate categories {A} and {B} V stoichiometric coefficient
specific area, m2/m3 vt turbulent kinematic viscosity, m2/s
EOD chemical oxygen demand, mol0, P density, kg/m3
c molar concentration, mol/m3 T shear stress, kg/( m s2)
CCOD corresponding concentration of chemical oxy- gas-phase backflow ratio
gen demand, COD/m3 : dimensionless radial coordinate
specific heat capacity, J/( kg K) * volume fraction in slurry
solid mass concentration in slurry, kg/m3
mean solid mass concentration, kg/m3 Indices
(axial) dispersion coefficient, m2/s abs absorption
axial dispersion coefficient, m*/s eff effective
radial dispersion coefficient, m*/s eq equations
reactor diameter, m C cell
gravitational acceleration, m/s* Cat catalysator
height, m C circulation
specific enthalpy, J/( kg) diss dissipation loss
jump factor [49] (feed) quantity in feed
volumetric mass transfer coefficient, s ’ G gas phase
 component index axial dispersion in gas-liquid cocurrent bubble columns, J.
mixing cell index Chem. Eng. Jpn., 20 (1987) 309-316.
k reaction index 20 H.-E. Gasche, Modellierung von BlasensPulen, Doctoral The-
sis, Univ. Erlangen, 1989.
L liquid phase
21 H.-E. Gasche, C. Edinger, H. Kompel and H. Hoffman,
R bubble column reactor Hydrodynamics in bubble columns, Chem. Eng. Process., 26
reaction (1989) lOl- 109.
ir) in radial direction 22 P. B. Whalley and J. F. Davidson, Liquid circulation in
S solid phase bubble columns, Inst. Chem. Eng. Symp. Ser. No. 38, (1974)
l-29.
T tube
23 J. B. Joshi and M. M. Sharma, A circulation cell model
V vortex for bubble columns, Trans. Inst. Chem. Eng., 57 (1979) 244
W heat transfer surface 251.
heat transfer surface index 24 J. B. Joshi, Axial mixing in multiphase contactors-a unified
;“x) in axial direction correlation, Trans. Inst. Chem. Eng., 58 (1980) 155-164.
25 P. Zehner, Impuls-, StoB- und Warmetransport in Blasen-
sHulen -Teil 1: Stromungsmodell der Blasensiiule und
Fliissigkeitsgeschwindigkeiten, Verfahrenstechnik, 16 (1982)
References 347-351.
26 P. Zehner, Impuls-, Stotf, und Wlrmetransport in Blasen-
saulen-Teil 2: Axiale Durchmischtmg und WPrmeiibergang,
1 W.-D. Deckwer, Reaktionstechnik in Blasensiiulen, Salle +
Verfahrenstechnik, 16(1982) 514-517.
SauerlZnder, Frankfurt/Main, 198.5.
27 S. Joseph and Y. T. Shah, A two bubble class model for chum
2 W.-D. Deckwer, Design and simulation of bubble column
turbulent bubble column slurry reactor, ACS Symp. Ser. No.
reactors, Proc. NATO Adv. Study Inst. “Chemical Reactor
237, 9(1983) 151-167.
Design and Technology”, London, Ont., Canada, 1985.
28 S. Joseph, Y. T. Shah, D. N. Smith and J. A. Ruether,
H. Gerstenberg, Blasensaulen-Reaktoren, Chem.-lng.-Tech.,
51 (1979) 208-216. Two-bubble class model for churnturbulent bubble column
reactor, Ind. Eng. Chem., Process Des. Deu., 24 (1985) 1096-
R. Steiner, Operation characteristics of special bubble column
1104.
reactors, Chem. Eng. Process., 21 (1987) l-8.
K. Weissermel and H.-J. Arpe, Industrial Organic Chemistry, 29 G. Grund. Hydrodynamische Parameter und Stoffaustauschei-
genschaften in Blasenslulen mit organischen Medien, Doctoral
VCH-Verlagsgesellschaft, New York, 1988.
H. J. Korte, Warmeiibergang in Blasensaulen mit und ohne Thesis, Univ. Oldenburg, 1988.
Einbauten, Doctoral Thesis, Univ. Dortmund, 1987. 30 M. PopoviE and W.-D. Deckwer, Berechnung axialdurchmis-
H. Ktirten, Verfahrenstechnik der Kohlehydrierung in Sumpf- chter Rohrreaktoren als Riihrkesselkaskade mit Rticklauf,
phasen-Reaktoren, Chem.-Ing.-Tech., 54 (1982) 409-415. Chem.-Ing.-Tech., 45 (1973) 984-990.
S. Ledakowicz and W.-D. Deckwer, Some aspects of design of 31 M. PopoviE, Eindimensionale Zellenmodelle fiir Gas-Fhissig-
commercial reactors for direct coal liquefaction? Chem. Eng. phase-Reaktoren mit axialer Durchmischung, Doctoral Thesis,
Commun., 25 (1984) 333-349. Tech. Univ. Berlin, 1975.
9 H. Kolbel and M. Ralek, The Fischer-Tropsch synthesis in 32 J. C. Mecklenburgh and S. Hartland, Design of reactors with
the liquid phase, Catal. Rev.-Sci. Eng., 21 (1980) 225-274. backmixing: Parts I- III, Chem. Eng. Sci., 23 (1968) 57-86.
10 J. C. W. Kuo et al., Slurry Fischer-Tropsch/Mobil-two 33 M. H. Roemer and L. D. Durbin, Transient response and
stage process of converting syngas to high octane gasoline, moments analysis of backflow cell model for flow systems with
Final Rep. US Dep. Energy, June 1983, Contract No. DOE- longitudinal mixing, Znd. Eng. Chem., Fundam., 6 (1967) 120-
PC-30022-10, Mobil Res. Dev. Corp., NJ. 129.
II D. Lindner and A. Schumpe, Zur Hydroxylamin-Herstellung 34 K. J. Myers, M. P. Dudukovic and P. A. Ramachandran,
nach dem HPO-Verfahren, Chem.-Ing.-Tech., 60 (1988) 302- Modelling churn-turbulent bubble columns-Part I: Liquid
304. phase mixing, Chem. Eng. Sci., 42 (1987) 2301-2311.
12 M. E. Frank and J. Klosek, Liquid phase methanol project 35 K. J. Myers, M. P. Dudukovii: and P. A. Ramachandran,
status, 6th Int. Symp. Alcohol Fuels Technology, Ottawa, Modelling liquid-phase chemical reaction and interphase mass
Canada, 1984, Vol. III, Cl 1, pp. 3-425-3-433. transfer in churn-turbulent bubble columns, Chem. Eng. Sci.,
13 M. Sherwin and D. Blum, Liquid-phase methanol, AF-1292 42 (1987) 2757-2766.
Res. Project 317-2, Final Rep. US Dep. Energy, Dec. 1989, 36 T. Wasowski and E. BlaB, Neue Aspekte zum EinfluB des
Electr. Power Res. Inst., Palo Alto, CA. Blasenwakesauf die Hydrodynamik und den Stoffaustausch in
14 R. Takahashi and K. Washimi, Steam crack vacuum residue, der Blasensaule, Chem.-Zng.-Tech., 61 (1989) 519-530.
Hydrocarbon Process., 55 ( 11) (1976) 93-96. 37 R. Torvik and H. F. Svendsen, Modeling of slurry reactors-
15 F.-J. Angenend, Die NaBluftoxydation von Schlamm; a fundamental approach, Chem. Eng. Sci., 45 (1990) 2325-
Brennst.- W&me-Kraft, 18 (1966) 252-258. 2332.
16 U. Fenchel, Klarschlammverwertung durch NaBoxidation 38 S. Schliiter, Modellierung und Simulation von Blasen-
nach Zimpro, Das Papier, 32 (lOA) (1978) VS4-V60. saulenreaktoren, Doctoral Thesis (submitted), Univ. Dort-
17 0. Levenspiel, Chemical Reaction Engineering, Wiley, New mund, 1992.
York, 2nd edn., 1972. 39 S. Wachi and H. Morikawa, Chlorination of ethylene in a
18 K. Ueyama and T. Miyauchi, Properties of recirculating tur- boiling bubble column reactor, J. Chem. Eng. Jpn., 20 (1987)
bulent two phase flow in gas bubble columns, AIChE J., 25 238-245.
(1979) 258-266. 40 A. Chianese, M. C. Annesini, R. De Santis and L. Marelli, A
19 S. Wachi, H. Morikawa and K. Ueyama, Gas holdup and model of a pilot bubble column reactor for benzene alkylation
 115

by ethylene, Chem. Eng. J., 22 (1981) 151-158. 56 A. A. C. M. Beenackers, G. H. Graaf, E. J. Stamhuis and N.
41 W.-D. Deckwer, Y. Serpemen, M. Ralek and B. Schmidt, P. Cheremisinoff, The synthesis of methanol, in N. P.,
Modeling the Fischer-Tropsch Synthesis in the slurry phase, Cheremisinoff (ed.), Handbook of Heat and Mass Transfer,
Ind. Eng. Chem., Process Des. Dev., 21 (1982) 231-241. Vol. 3: Catalysis, Kinetics and Reactor Engineering, Gulf,
42 D. M. Brown et al., Modeling of methanol synthesis in the liquid Houston, TX, 1989, pp. 671-700.
phase, ACS Symp. Ser. No. 87, (1983) 699-709. 57 M. B. Sherwin and M. E. Frank, Make methanol by three
43 D. Stern, A. T. Bell and H. Heinemann, A theoretical model phase reaction, Hydrocarbon Process., 55 ( 11) (1976) 122-
for the performance of bubble column reactors used for Fis- 124.
cher-Tropsch synthesis, Chem. Eng. Sri., 40( 1985) 1665-1677. 58 L. W. Bonnel and R. F. Weimer, Slurry reactor design for
44 A. Schumpe, Y. Serpemen and W.-D. Deckwer, Effective methanol production, AIChE Annu. Meeting, San Francisco,
application of bubble columns, Ger. Chem. Eng., 2 (1979) CA, USA, 1984.
234-241. 59 W. Kotowski et al., Synthese von Alkohol-Gemischen an
45 S. S. &irk, Y. T. Shah and W.-D. Deckwer, Comparison of suspendierten Kontakten, Chem.-Ing.-Tech., 58 (1986) SlO-
gas and liquid phase methanol synthesis processes, Chem. Eng. 812.
J., 37 (1988) 177-192. 60 W. von Wedel, S. Ledakowicz and W.-D. Deckwer, Kinetics
46 J. R. Turner and P. L. Mills, Comparison of axial dispersion of methanol synthesis in the slurry phase, Chem. Eng. Sci., 43
and mixing cell models for design and simulation of Fischer- (1988) 2169-2174.
Tropsch slurry bubble column reactors, Chem. Eng. Sci., 45 61 G. H. Graaf, P. J. J. M. Sijtsema, E. J. Stamhuis and G. E. H.
(1990) 2317-2324. Jo&en, Chemical Equilibria in methanol synthesis, Chem. Eng.
47 M. Kralik and J. Ilavsk$, A mathematical dynamic model of Sci., 41 (1986) 2883-2890.
the continuous bubble column slurry reactor, Chem. Eng. 62 G. H. Graaf, J. G. M. Winkelman, E. J. Stamhuis and A. A.
Process., 28 (1990) 127-132. C. M. Beenackers, Kinetics of the three phase methanol syn-
48 J. B. Joshi and M. M. Sharma, Heat transfer in multiphase thesis, Chem. Eng. Sci., 43 (1988) 2161-2168.
contactors, Chem. Eng. Commun., 6 (1980) 257-271. 63 S. Ledakowicz, Kinetics of the slurry phase methanol synthe-
49 D. N. Smith and J. A. Ruether, Dispersed solid dynamics in a sis, Proc. 4th World Cong. Chemical Engineering, Karlsruhe,
slurry bubble column, Chem. Eng. Sci., 40 (1985) 741-745. Germany, 1991.
50 VDI- W&meat/as- Berechnungsbliitter ftir den Wiirmeiiber- 64 W. von Wedel et al., Methanol-Synthese am suspendierten
gang, VDI-Verlag, Diisseldorf, 6th edn.: 199 1. Kontakt-LBslichkeitsdaten, Chem.-Ing.-Tech., 60 (1988)
51 R. A. Novv. H. T. Davis and L. E. Striven. Uastream and 492-493.
downstream boundary conditions for continuohs flow sys- 65 H. Perkov, R. Steiner and H. Vollmiiller, Wet air oxidation-
tems, Chem. Eng. Sci., 4.5 (1990) 15 I5- 1524. a review, Ger. Chem. Eng., 4 (1981) 193-201.
52 P. V. Danckwerts, Continuous flow systems-distribution of 66 R. Randolf, NaDluftoxydation von Abwasserschlamm,
residence times, Chem. Eng. Sci., 2 (1953) 1-13. Wasserwirtschr.- Wassertech., 18 (1968) 342-346.
53 NAG-The Numerical Algorithm Group, The NAG-FOR- 67 J. J. A. Ploos van Amstel and K. Rietema, NaBoxidation von
TRAN Library, Version 14.0, Numerical Algorithm Group, Abwasserschlamm, Teil I: Oxidation von Glucose als Modell-
London, 1991. substanz, Chem.-lng.-Tech., 42 (1970) 981-990.
54 N. Kijckler, Numerische Algorithmen in Softwaresystemen, 68 J. J. A. Ploos van Amstel and K. Rietema, Wet-air oxidation
Teubner, Stuttgart, 1990. of sewage sludge, Part II: The oxidation of real sludges,
55 Ph. Rabinowitz (ed.), Numerical Methods for Nonlinear Alge- Chem.-Ing.-Tech., 45 ( 1973) 1205.~ 1211,
braic Equations, Gordon and Breach, London, 1988. 69 J. Beyrich, NaDoxidation, Tech. Mitt., 80 (1987) 350-352.

Appendix overleaf
 Appendix

(l&h)(;)
+

- ,$Nl
I

cell (N) cell (N)

Q;“’

$N-‘1 (1 + (‘“-1)) (1 + 7(N-‘)) (&L)(N)

(&)‘“-‘)7(h.-‘l (1 t P-
II cl + 7(j)) ( $‘pph)‘;l
(1 t 76)) (tit,,.,)“’ #) ($&)(I+‘) 7’j) (eh)‘;+”

k(j)
.I

cell (j) p2 cell (j)

N(j)
I&

+ +

(7t 7(l)) (I+ 7~)) (t+phf 7(” (Iip$

Fig. A-l. Cell model with backflow: overall mass balance, gas phase (left-hand side) Fig. A-2. Cell model with backflow: mass balance component i, liquid phase (left-hand
and mass balance component i, gas phase (right-hand side). side) and energy balance, gas and liquid phases (right-hand side) for the case of
cocurrent gas/liquid flow operation.
+
s